Idle apps stay cheap
CPU bills only while it runs. A metered workload scaled to 0 replicas stops metering entirely; you pay only for the memory a running app holds.
Cost optimization
Ownkube meters per minute for what you actually run, so keeping costs down is mostly about running lean: let idle apps idle, right-size your requests, and cap your budget. This page covers the levers on Ownkube-hosted compute, then the free Starter path if you run in your own cloud.
The cost model in one paragraph
Section titled “The cost model in one paragraph”On Ownkube-hosted compute you draw a prepaid wallet: CPU meters on actual per-minute usage and memory on the working set your app holds, at about $0.00041 / vCPU-minute and $0.0001 / GiB-minute. A mostly-idle app barely registers. That is roughly a third under the equivalent Railway bill. See Pricing for the full breakdown.
Where the savings come from
Section titled “Where the savings come from”Scale without waste
Horizontal autoscaling adds replicas under load and removes them when traffic drops, so you only pay for the capacity you’re actually using. See Autoscaling.
Right-size your requests
Set honest CPU and memory limits per workload. They are ceilings, not a fixed price, so the bill follows the meter, not the size you pick.
Reserved boxes for steady load
A workload that runs flat-out all month can opt into a reserved box for a predictable price instead of the meter. Both draw the same wallet. See Pricing.
Cap your spend
Set a monthly budget and choose to be notified or to have apps pause at the cap. Spending alerts email you at 50%, 80%, and 100% of the budget, so a bill can’t run away. See Spend limits.
Skip the $200K DevOps hire
A US senior DevOps engineer runs $180K to $250K loaded. Automatic TLS, managed backups, and autoscaling cover the recurring platform work day to day, so a 5- to 20-person team can stay focused on product.
A worked example
Section titled “A worked example”A small app (one web service at 0.5 vCPU / 512 MiB and one worker at 0.25 vCPU / 256 MiB) that stays mostly idle, plus a small managed Postgres (reserved at 1 vCPU / 1 GiB):
| Component | How it meters | Monthly cost |
|---|---|---|
| Web service | Metered, mostly idle | ~$3 to 5 |
| Worker | Metered, mostly idle | ~$2 to 3 |
| Postgres (reserved 1 vCPU / 1 GiB) | Reserved, continuous | ~$22 |
| Egress (modest) | First 100 GB free | $0 |
| Total | ~$27 to 30 |
The database, billed reserved, is the bulk of the bill; the app itself is a few dollars while it sits idle. Right-size the database and you move the whole number.
On the roadmap
Section titled “On the roadmap”In-dashboard cost visibility is on the roadmap. Planned capabilities:
- Auto-sleep schedules (roadmap) for non-production environments during off-hours. Example: “3 staging envs idle since 8pm. Auto-slept. ~$127/mo saved this week.”
- Resource sizing recommendations (roadmap) to flag deployments with chronically low utilization against what they’ve requested. Example: “api-worker is provisioned for 2GB but peaks at 340MB. Lowering the request could save about $18/mo.”
- Cost anomaly alerts (roadmap) when spend jumps outside its normal band. Example: “data-transfer-out 4x higher than 7-day baseline. Investigate before month-end.”
Running in your own cloud
Section titled “Running in your own cloud”If you run in your own AWS account, the Starter shape (K3s on one AWS instance) has no Ownkube platform fee: you pay AWS directly for the instance (from ~$30/month on t3.medium) and can run unlimited apps, workers, and databases on it. Adding another service to the same node is $0 extra. Production (managed EKS) adds a $5 / vCPU + $1 / GiB RAM platform fee, metered on requested resources. See Pricing.
Your AWS bill is the canonical source of truth for compute cost in that mode, and any Reserved Instances or Savings Plans you hold apply directly.
Pricing Full rate card, reserved boxes, and spend limits.
Autoscaling Scale on CPU and memory so you only pay for load you actually get.